This assignment is for ETC5521 Assignment 1 by Team Echidna comprising of Ruimin Lin, Rahul Bharadwaj, Ketan Kabu, and Yezi He.
Board Game has been a type of leisure that people have enjoyed from a very long time even before computers and video-games existed and has gone through enormous evolution ever since its inception. Board Games enables a way for people to socialize, reducing stress under such a fast-moving society, and paves way for an extensive brain exercise. Being a popular choice of leisure, what makes board games great? What is the reason for Board Games to have survived in a world of Virtual Reality games? In other words, what are the common characteristics of top ranked board games? What are the best board games in terms of average rating?
The original board games data used in this report is obtained from the Board Game Geek database, and is cleaned and shared by Thomas Mock.
The tidy dataset consists of 22 columns and 10532 rows, in which there are 22 variables and 10532 observations. It consists of data such as max/min playtime, max/min players, min age of players that can play, game designer, game publisher, mechanics of the game and a lot more. One thing to notice is that even though the data set is tidy, we still find observations in variables like category, family, mechanic to be messy and repetitive, which may limit our ability to explore these variables.
The aim of this exploratory analysis is to find out what factor affects the average rating of board games. This would give insights as to what board games are most popular and the characteristics these board games share. Therefore, we have articulated the following questions to help us with further exploration of the board games data.
Primary Question:
What are the common characteristics of top ranked board games?
Secondary Questions:
The variables included in the data are as follows:
game_id: ID of a particular game, the game_id should be a character vector(categorical) instead of a double vector mentioned in the table above.
description: Game description, a character vector.
image: URL image of the game, a character vector.
max_players/min_player: maximum/minimum number of recommended players, double vectors.
max_playtime/min_playtime: maximum/minimum recommended playtime, double vectors.
min_age: recommended minimum player age, double vectors.
name: name of the game, a character vector.
playing_time: average playtime of a game, a double vector.
thumbnail: URL thumbnail of the game, a character vector.
year_published: year the game was published, a double vector.
artist: artist for game art, a character vector.
category: categories of the game, a character vector.
compilation: name of compilation, a character vector.
designer: game designer, a character vector.
expansion: name of expansion pack (if any), a character vector.
family: family of game - equivalent to a publisher, a character vector.
mechanic: how game is played, a character vector.
publisher: company/person who published the game, a character vector.
average_rating: average rating from 1 to 10 on the website(Board Games Geek), a double vector.
users_rated: number of users rated the game, a double vector.
To ensure the reliability of the board game ratings, the data is limited to games with at least 50 ratings and for games between 1950 and 2016. The site’s database has more than 90,000 games with crowd-sourced ratings.
The original board games data set consists of 90400 observations, and 80 variables. Therefore, data cleaning and wrangling is necessary to enable better analysis procedure. Thomas has replaced long and complicated variable names like details.description in original data to description using janitor::clean_names and set_names function, which avoids messy code writing. In addition, he has eliminated around 50 variables using the select function and that leaves 27 variables at this stage.
The data set is then filtered to board games published from 1950 to 2016, with at least 50 users rated. ‘NA’ values in variable year_published is also omitted. Thomas then excludes variables that may not be useful for the analysis, such as attributes_total, game_type etc., which ultimately, leaves us with a tidy data set (22 variables and 10532 variables) that is relatively concise and convenient for further exploration.
Figure 3.1: Visualization of Data Types
designer and publisher columns for our analysis.Figure 3.2: Visualization of Missing Values
The above plot 3.2 shows the percentage of missing values and where exactly they are missing with x-axis showing columns and the y-axis showing the corresponding observations. We can also observe that each column has a percentage of missing values mentioned which come in handy while deciding what columns not to pick for analysis.
It is evident that the following columns have missing values and are not of much use for the analysis:
compilation - 96.11% missingexpansion - 73.87% missingfamily - 26.66% missingmechanic - 9.02% missingThis is a limitation of the dataset and we frame our questions keeping this in mind.
Figure 3.3: Top 10 ranked board games
| Game | Average rating | Max playtime | Min playtime | Max players | Min players |
|---|---|---|---|---|---|
| Small World Designer Edition | 9.00 | 80 | 40 | 6 | 2 |
| Kingdom Death: Monster | 8.93 | 180 | 60 | 6 | 1 |
| Terra Mystica: Big Box | 8.85 | 150 | 60 | 5 | 2 |
| Last Chance for Victory | 8.85 | 60 | 60 | 2 | 2 |
| The Greatest Day: Sword, Juno, and Gold Beaches | 8.83 | 6000 | 60 | 8 | 2 |
| Last Blitzkrieg | 8.80 | 960 | 180 | 4 | 2 |
| Enemy Action: Ardennes | 8.76 | 600 | 0 | 2 | 1 |
| Through the Ages: A New Story of Civilization | 8.74 | 240 | 180 | 4 | 2 |
| 1817 | 8.71 | 540 | 360 | 7 | 3 |
| Pandemic Legacy: Season 1 | 8.67 | 60 | 60 | 4 | 2 |
Figure 3.4: Visualization of Data Types in Top 50 Games
The above plot 3.4 shows a distribution of Data Types in our Top 50 Games dataset with x-axis showing column names and y-axis its corresponding observations.
It is evident that our selection of columns is appropriate and there are no missing values in our data. Hence, we need not check for missing values through vis_miss() function. We can use all these columns for an effective analysis of our questions of interest.
Figure 3.5: Relationship between Maximum Playtime and Average Rating
To have a better idea on the common characteristics of top-ranked board games and ensuring the reliability of the results, we have widened the range to top 50.
In plot 3.5 we can see that there are a few obvious distinct values present, which are:
It is difficult to examine the trend or common characteristics with these outliers presents, therefore, we have limited the maximum playtime to less than xx minutes using the IQR outliers formula. (Q1 - 1.5IQR and Q3 + 1.5 IQR)
## # A tibble: 1 x 6
## minimum q1 median mean q3 maximum
## <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 0 82.5 142. 461. 345 6000
## # A tibble: 1 x 2
## lower_range upper_range
## <dbl> <dbl>
## 1 -311. 739.
Figure 3.6: Adjusted Maximum Playtime
Now we can have a clearer picture of where majority of top-50 ranked board games lie in the graph of average rating against maximum playtime. Which, majority of board games lie within the range of 200 minutes of maximum playtime, the highest ranting board game also lies within the range, around 100 minutes of maximum playtime. Another thing to notice is that, for board games that have maximum playtime longer than 600 minutes, the rating is comparatively lower.
Nearly half of high rating board games are crowded in the range of 0-200 minutes, suggesting that people tend to play board games that does not occupy too much leisure time.
Figure 3.7: Minimum Playtime Plot
Figure 3.8: Relationship between Minimum Players and Average Rating
Figure 3.9: Relationship between Maximum Players and Average Rating
In the scatterplot for average rating against maximum players, we observed that most top 50 board games have a maximum of 4 or 5 players.
The figure 3.8 and 3.9 indicates that majority of high rating board games have set the players to between 2 and 4/5 players. The limitation of players suggest that people tend to play board games that fulfills their sense of participation, for example, a board game of 8 players may not be as attractive as a board game of 2 players, because a 2-player game has little downtime than a 8-player game, and satisfies each players’ sense of participation in the board game.
On the other hand, it is easier to gather a group of 2-4 people interesting in play board games at leisure time than gathering a group of 8 or more people.
Figure 3.10: Minimum Age Plot
Figure 3.11: Summarizing all observations as Violinplots
Figure 3.12: Relationship between Average Rating and other Attributes
The above plot 3.12 shows a trend for different attributes against average rating on x-axis. We can get a better idea using this pattern.
We can observe the following trend for the top 50 rated games as average rating increases -
Figure 3.13: Minimum Age Plots
We can observe the following for the attribute Minimum Age -
Important note - Majority of the games are classified into multiple categories at once. For example, the game Top Trumps has about 14 categories it has been classified under. Therefore, the calculation of number of games under each category will be an overlapping process, as one game would be classsified under various categories. If one would calculate the percentage of games under all category, it will be higher than 100%.
Even so, it gives us a good understanding about the popularity of a category, and for new game developers to look into which category they should target for their next upcoming games.
Figure 3.14: Top 10 categories of board games in 2016. The length of the bars parallel to the x-axis represent average rating of all categories, with each of the the top 10 categories mapped on the y-axis. Numbers on the right of bars represent games of those categories launced in 2016. Clearly observable that war type games are most liked, with 43 games in Wargame categories in 2016 alone.
Numbers on the right hand side of bars represent the number of games of that category that were published in 2016. This gives us a better understanding of the ratings as some will be spread out over a number of games, even if they are relatively lower in ranking, like War Games and Miniatures categories. They have 43 and 29 games published respectively in 2016, and yet have maintained a top 10 ranking of 6 and 8 respectively. Vitenam War category has just one game published but was rated quite highly (7.75). American Indian Wars category had 2 games published in 2016, and was the highest rated category with an average rating of 7.85
It is clear from the above graph that American Indian Wars category has the highest rating in the most recent yearly data available (2016).
An interesting point to note about human behavior and how we like to pass our free time, when we see that the top 7 out of those top 10 categories of board games are war related.
For anyone reading this paper in a relevant field, it would be an interesting psychological experiment to understand why we, as a species, are so drawn to war-like scenarios even during our innocent board game time.
Additionally, a new game developer could draw an insight from this chart that war based games are the popular, when brainstorming ideas for new games to launch.
Let us now consider, how these categories in figure 3.14 above have evolved over time. We can take a look at a timeline of all these board games and see if we can observe any trends in the likability or acceptance of board game categories.
Figure 3.15: Current top 10 board games rating over the last 50+ years. Decade and Rating mapped on x and y axis resepctively. Numbers on graph represent number of games published each decade in that category. Each facet represents a game category in 2016 top 10. A steep rise in likability since last decade is visible for all games, with a dip for nearly all categories in the 1980s. Important to consider that number of games in 2010s are only till 2016.
An initial look at the figure 3.15 above indicates that nearly all these 2016 top 10 categories have undergone significant rise in ratings in the last decade.
Nearly all war themed games (except Modern Warfare) had undergone a drop in ratings in the 1980s, and then sharply rose and kept rising till 2016. World War did not see a drop in ratings in the 80s, but had it’s growth plateaued as well.
Vietnam war (the actual war) lasted 2 decades from 1950s to the 70s. The first game introduced in this category had a rating just above 7 around the time the war ended. Ratings droppped as with most war games in the 1980s, and then rose again to be the second highest rated category as of 2016.
When we look at the number of games manufactured for each category and each decade, it is of note that the top rated category in 2016 - American Idian wars, had only 1 game in the 1990s, when it dropped to it’s lowest point. This was followed by 2 and then 6 games in the following decades, when it rose to the top of the rankings.
There were 545 games of the Wargame category published in 2000s, followed by a drop in numbers to 424 in 2010s, but ratings rose to nearly 7.5.
Figure 3.16: Top 10 Game Designers
The above plot 3.16 consists of average rating on x-axis and designer on y-axis. The black x-intercept represents the mean value of average ratings of the top 10 designers. The plot conveys that the mean average rating is around 8.82 with 5 observations on either side of the line.
Philippe Keyaerts has the highest rated game at around 9+ followed by Vlaada Chvatil around 8.93 with all the other designers falling around the mean value. The lesser rated designer in the top 10 is Rob Daviau, Matt Leacock. We should note that Dean Essig has two games in the top 10.
Who among these is the best is still a debatable question. Some might say it is Dean, while some might consider Philippe. Nevertheless, all of the designers in the plot are among the top 10 and have produced the most popular games.
Figure 3.17: Top 7 Game Publishers
The above scatter-plot 3.17 consists of average rating on x-axis and publisher on y-axis.
The first thing that strikes from looking at this plot is that Multi-Man Publishing has 3 among the top 7 rated board games which hints that they are one of the best publishers.
The top rated game was published by Days of Wonder and the lesser rated game in the top 7 was published by Compass Games. Again, it is debatable as to who is best but the publishers in the above plot have published some of the finest board games.
Bonus Insight - An interesting takeaway from the above two plots is that the best and top rated board games were launched between 2010-2016 with most of the top rated games launched in the year 2015.
Figure 3.18: Top 10 Mechanic with rating
The treemap 3.18 shows the top 10 frequent mechanic type, the size of rectangles corresponding to the frequency of mechanic.
The most popular mechanic is Hex-and-Counter with a 6.8 average rating. Among the top 10 mechanic, Dice Rolling,Hex-and-Counter,Simulation rated highest. The lesser rated mechanic in the top 10 is 4 Roll / Spin and Move.
In the top 10 blue rectangles, that is, machines with an average score higher than 6.6, all contain Hex-and-Counter.
From the above analysis, we can see that the top ranked board games have the following common characteristics. A maximum of 4 players and minimum of 2 players is most popular in the top 50 games. The maximum and minimum playtime seem to be almost close and range between 60-150 minutes for top 50 games. The minimum age set by majority of board games are between 10 - 15. American Indian Wars category has the highest rating in the most recent yearly data available. Philippe Keyaerts has the highest rated game and Multi-Man Publishing is one of the best publishers. Hex-and-Counter is the most popular mechanic.
Websites
BoardGameGeek | Gaming Unplugged Since 2000. (2000). BGG. https://boardgamegeek.com/BoardGameGeek
Huebner, M., Vach, W. and le Cessie, S., 2016. A systematic approach to initial data analysis is good research practice. The Journal of Thoracic and Cardiovascular Surgery, 151(1), pp.25-27.
Thomas Mock, (2019). Tidy Tuesday. https://github.com/rfordatascience/tidytuesday/tree/master/data/2019/2019-03-12
R packages
Baptiste Auguie (2017). gridExtra: Miscellaneous Functions for “Grid” Graphics. R package version 2.3. https://CRAN.R-project.org/package=gridExtra
C. Sievert. Interactive Web-Based Data Visualization with R, plotly, and shiny. Chapman and Hall/CRC Florida, 2020.
David Robinson, Alex Hayes and Simon Couch (2020). broom: Convert Statistical Objects into Tidy Tibbles. R package version 0.7.0. https://CRAN.R-project.org/package=broom
Hao Zhu (2019). kableExtra: Construct Complex Table with ‘kable’ and Pipe Syntax. R package version 1.1.0. https://CRAN.R-project.org/package=kableExtra
Tierney N (2017). “visdat: Visualising Whole Data Frames.” JOSS, 2(16), 355. doi: 10.21105/joss.00355 (URL: https://doi.org/10.21105/joss.00355), <URL: http://dx.doi.org/10.21105/joss.00355>.
Wickham et al., (2019). Welcome to the tidyverse. Journal of Open Source Software, 4(43), 1686, https://doi.org/10.21105/joss.01686
Yihui Xie (2020). bookdown: Authoring Books and Technical Documents with R Markdown. R package version 0.20.
Yihui Xie (2020). knitr: A General-Purpose Package for Dynamic Report Generation in R. R package version 1.29.
Martijn Tennekes (2017). treemap: Treemap Visualization. R package version 2.4-2. https://CRAN.R-project.org/package=treemap
Kamil Slowikowski (2020). ggrepel: Automatically Position Non-Overlapping Text Labels with ‘ggplot2’. R package version 0.8.2. https://CRAN.R-project.org/package=ggrepel